Catching defects before they ship
Line-side cameras that inspect every part, flag defects in real time and show where quality is slipping.
- Type
- Solution showcase
- Industry
- Manufacturing
- Deployment
- Line-side camera and edge GPU, linked to the PLC
- Services
- Computer Vision, AI & ML
The challenge
Human inspectors are skilled, but they tire, and sampling one part in fifty lets defects through. When a customer finds the problem first, the cost is far higher than the part.
The solution
Inspect AI places a high-resolution camera at a station on the line. A model classifies every part as good or defective, identifies the defect type, such as a scratch, dent or missing component, and signals the line to divert rejects. Quality teams see defect rates by type, shift and machine, and new defect types are labelled and added to the model as they appear.
What it includes
- Inspection of every part, not a sample
- Defect classification by type
- Reject signal to the line controller
- Defect dashboard by shift and machine
- Review queue for uncertain parts
How it works
- 01
Image
A camera with controlled lighting captures every part at a fixed station.
- 02
Classify
A model trained on your good and defective parts makes a decision in milliseconds.
- 03
Divert
Rejects are signalled to the line controller and set aside for review.
- 04
Improve
Quality teams review edge cases, and the model is retrained on new examples.
Design targets
This is a solution showcase: a system Quibo designs and builds for this kind of problem. Figures are the design targets we build and test against, not results from a named client. Targets for your business are agreed and measured during the proof stage.
Services used
Want a system like this?
Tell us about the process or the cameras you have. We will tell you honestly what is achievable, and propose a short proof stage on your own data.
Start a conversation